This semester, I’m the TA for PSY 330, a beginner’s guide to computational cognitive neuroscience — that is, using biologically realistic neural network models to explain cognitive phenomena. You have to learn a bit about network dynamics to do this properly, and so my basins around attractors like, well, attractors have deepened considerably since February (see upcoming novel excerpt for further evidence). I also recently read Malcolm Gladwell’s book OUTLIERS, which makes a substantial set of the same points as Dick Nisbett’s book INTELLIGENCE AND HOW TO GET IT, namely that talent is usually a result of small native predispositions amplified by environmental factors. Nisbett’s example is basketball: A kid who’s slightly tall tends to get picked first for basketball teams, tends to do better when he plays, tends to get more praise from peers and superiors — and by virtue of this, ends up better at basketball than someone who’s slightly short, because he plays a lot more basketball. (Nisbett’s very apt point in bringing this up is that we might, if we were unreflective, be inclined to attribute this considerable variance to genetics, when most of it was due to environmental amplification; Gladwell is more comfortable assuming that runaway talents, at least, tend to have commanding genetic advantages, although those don’t suffice for success in the absence of environmental amplifiers.)
This is an example of positive feedback, which is what makes microphones screech when you point them at the speaker they’re attached to; it’s also the way units in a randomly initialized network develop selectivity (units whose weights happen to correspond to some meaningful input feature win a little bit over inhibition, get strengthened, win even more, etc.), which is never described as positive feedback, but I guess it is. So the question is, what accounts for the difference between a runaway talent and a talent who’s just very good?
The question is interesting not only for obvious reasons (how do I become the next Kanye West?), but also because it gets you focused on what exactly is being fed back. To my eye, there are three things: initial level (innate and ex hypothesi immutable), reinforcement for the behavior (in the form of praise, satisfaction, money, reputation), and opportunity to do more of it (in the form of invitations to perform or professionalization of performance). These things are substantially although not totally independent — the anhedonic Olympian gymnast type gets a lot of opportunity but not a lot of reward, the streetball sensation type gets a lot of reward but not a lot of opportunity. Of course, they both have a substantial amount of each — you can’t practice without opportunity, and you won’t without reward, or at least the avoidance of punishment. But the point is that sensitivity to either variable can affect the rate of feedback — equivalently, but framed in a particularly trenchant investment metaphor, the rate of compounding.
And this is interesting because it suggests ways to succeed that don’t involve social engineering to equalize opportunity, which is what Gladwell and Nisbett advocate — correctly, of course, but impracticably for most of us on an individual level. If you can learn to take pleasure from improvement and scale your practice proportional to improvement, you will ride the curve up; the bigger those scaling factors, the faster you go. These are not easy things, but they are at least under your control. And perhaps the most interesting thing is that, to ride the curve, you have to know how good you are.
This is all out of first principles and secondhand knowledge. Is there any independent reason to believe that extremely talented people have a very precise sense of their own level?